System and Method for Deep Learning-Based Chemical Shift Artifact Mitigation of Non-Cartesian Magnetic Resonance Imaging Data

1. A computer-implemented method for generating a chemical shift artifact corrected reconstructed image from magnetic resonance imaging (MRI) data, comprising: inputting into a trained deep neural network an image generated from the MRI data acquired during a non-Cartesian MRI scan of a subject; utilizing the trained deep neural network to generate the chemical shift artifact corrected reconstructed image from the image, wherein the trained deep neural network was trained utilizing a tissue mixing model that models interactions between different tissue types to mitigate chemical shift artifacts, and wherein the tissue mixing model comprises a partial volume map for approximating a respective fraction of the different tissue types in each voxel of the image; and outputting from the trained deep neural network the chemical shift artifact corrected reconstructed image.

2. The computer-implemented method of claim 1, wherein the MRI data is acquired at a lower receiver bandwidth, and the chemical shift artifact corrected reconstructed image appears as generated from MRI data acquired at a higher receiver bandwidth, wherein the higher receiver bandwidth is greater than the lower receiver bandwidth.

3. The computer-implemented method of claim 1, further comprising training a neural network using supervised learning to generate the trained deep neural network, wherein training data used for the supervised learning comprises original images without chemical shift artifacts and corresponding images with chemical shift artifacts retrospectively generated from the original images, wherein the original images function as a ground truth.

4. The computer-implemented method of claim 1, wherein the tissue mixing model is configured to simulate phase accrual of the different tissue types with known chemical shift evolution.

5. The computer-implemented method of claim 1, further comprising receiving an input of a user selection of either fully removing or partially removing a chemical shift artifact in the image, wherein, when the input of the user selection is for fully removing the chemical shift artifact, the chemical shift artifact is fully removed from the chemical shift artifact corrected reconstructed image, and wherein, when the input of the user selection is for partially removing the chemical shift artifact, the chemical shift artifact is partially removed from the chemical shift artifact corrected reconstructed image.

6. The computer-implemented method of claim 1, wherein the image is the only image inputted into the trained deep neural network to generate the chemical shift artifact corrected reconstructed image.

7. The computer-implemented method of claim 1, wherein the non-Cartesian MRI scan comprises an on-resonance scan.

8. The computer-implemented method of claim 1, wherein the non-Cartesian MRI scan comprises an off-resonance scan.

9. The computer-implemented method of claim 1, further comprising inputting physics-based information into the trained deep neural network, wherein the trained deep neural network utilizes the physics-based information to guide correction in generating the chemical shift artifact corrected reconstructed image.

10. The computer-implemented method of claim 9, wherein the physics-based information comprises one or more MRI images at different frequencies generated from the original MRI data.

11. The computer-implemented method of claim 9, wherein the physics-based information comprises a map of a primary magnetic field of an MRI system utilized to acquire the MRI data.

12. The computer-implemented method of claim 9, wherein the physics-based information comprises scalar information.

13. The computer-implemented method of claim 1, wherein the image and the chemical shift artifact corrected reconstructed image are two-dimensional images.

14. The computer-implemented method of claim 1, wherein the image and the chemical shift artifact corrected reconstructed image are three-dimensional images.

15. A non-transitory computer-readable medium, the computer-readable medium comprising processor-executable code that when executed by a processor, causes the processor to: input into a trained deep neural network an image generated from magnetic resonance imaging (MRI) data acquired during a non-Cartesian MRI scan of a subject; utilize the trained deep neural network to generate a chemical shift artifact corrected data from the image, wherein the trained deep neural network was trained utilizing a tissue mixing model that models interactions between different tissue types to mitigate chemical shift artifacts, and wherein the tissue mixing model comprises a partial volume map for approximating a respective fraction of the different tissue types in each voxel of the image; and output from the trained deep neural network the chemical shift artifact corrected data.

16. A deep learning-based chemical shift artifact correction system for generating a chemical shift artifact corrected data from magnetic resonance imaging (MRI) data, comprising: a memory encoding processor-executable routines; a processor configured to access the memory and to execute the processor-executable routines, wherein the routines, when executed by the processor, cause the processor to: input into a trained deep neural network an image generated from the MRI data acquired during a non-Cartesian MRI scan of a subject; utilize the trained deep neural network to generate the chemical shift artifact corrected data from the image, wherein the trained deep neural network was trained utilizing a tissue mixing model that models interactions between different tissue types to mitigate chemical shift artifacts, and wherein the tissue mixing model comprises a partial volume map for approximating a respective fraction of the different tissue types in each voxel of the image; and output from the trained deep neural network the chemical shift artifact corrected data.

17. The system of claim 16, wherein the routines, when executed by the processor, cause the processor to input physics-based information into the trained deep neural network, wherein the trained deep neural network is configured to utilize the physics-based information to guide correction in generating the chemical shift artifact corrected data.